Using limited common sense knowledge to guide knowledge acquisition for information agents

Acquiring knowledge from the Internet to improve performance is a very challenging task for information agents. This research aims to use common sense knowledge to guide the learning. A limited common sense knowledge base is handcrafted into an agent using extended Definite Clause Grammar (DCG) rules. A Classified Advertisement Search Agent (CASA) reading advertisements in HTML has been built and has been tested on real estate advertisements. CASA learns new knowledge about suburbs automatically, and acquires statistical information about prices that can be used for making suggestions to modify user queries, and to fill in default parameters.